Re: Help Deciphering an Example

[Hans Aberg]

On 4 Oct 2005, at 22:48, Jason Ki wrote:

> This is in regards to the thread "token value return in custom yylex 
> ()"
> posted by Bernd Prager on Tuesday December 3rd, 2002.
>
> http://lists.gnu.org/archive/html/help-bison/2002-12/msg00002.html
>
> His grammar code shows the following:
>
> ----- start of example -----
> %{
> #include <stdio.h>
> #include <string.h>
> #include "parser.tab.h"
>
> #define TESTSTRING "word1 word2" // testpattern
>
> char *data, buffer[1024];
> %}
>
> %union {
>     int id;
>     char *str;
> }
>
> %type <str> object token list
> %token <id> WORD
>
> ----- end of example -----
>
> I am unsure what the previous coding really means. At first I thought
> that a token could either be a string or an integer. However, on a
> closer inspection I noticed that the it was "%type" and not "% 
> token". So
> does this mean that a token can only be of type integer? If this is
> wrong, which I'm positively sure it is, then I believe I still have  
> yet
> to grasp what a token is.

Bison has a static typing system, which is turned on when %union is  
invoked, and at the same time the stuff in the %union clause is  
implemented as a C-union. One then has to figure out how to select  
the right union field, which is done with the %type and %token  
declarations above. Each grammar symbol can only have one type value,  
and that is written down above. It says that the token WORD has the  
type <id>, the grammar variables object, token, list, have type <str>.

> Additionally, is yacc similar to BISON? The website below:
> http://epaperpress.com/lexandyacc/index.html
> implies that yacc is the same as BISON in regards that both are  
> software
> used to generate grammar rules for building a syntax tree.

There is a POSIX specification for what a Yacc-like program should  
fulfill, and Bison has a mode where it aims to compliant with that  
POSIX specification.

  Hans Aberg